[Rasch] [Semi-OT] How to simulate data with error
rense.lange at gmail.com
Fri Jun 28 11:28:49 EST 2013
TYPO: There is NO obviously right answer ….
On Jun 27, 2013, at 10:25 PM, Rense Lange <rense.lange at gmail.com> wrote:
> There is obviously right answer as the conditional SE increase away from the middle of the test. And, the rate of increase varies with the distribution of item difficulties and the number of items - as you probably well know. Given those two things, you can randomly select a measure (say, V) and then draw from a normal distribution N(V,SE).
> But, somehow you will have to start with reasonable (and intended) V and SE, and those are non arbitrary as V and SE vary with the items' difficulty.
> On Jun 27, 2013, at 10:09 PM, Stuart Luppescu <slu at ccsr.uchicago.edu> wrote:
>> al response data, just the measures.
>> I'm interested in looking at the effect of different amounts of error in
>> the measures on the results of the analysis. So, I'm thinking I would
>> add different amounts of random noise to the measures as a way of
>> introducing varying error. This is a sort of CCT-ish way of thinking but
>> how else can I include different amo
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Rasch